Why are companies spending heavily on AI tools while struggling to show meaningful improvements in productivity, revenue, or business performance?

In this episode of Tech Talks Daily, I speak with Matt Cloke, Chief Technology Officer at Endava, about what it takes to become an AI-native business, why deploying thousands of AI licenses does not amount to an AI transformation, and how companies can move from experimentation to measurable business outcomes.

Matt has played a central role in Endava's own adoption of artificial intelligence and the development of Dava.Flow, the company's methodology for applying AI throughout the technology delivery lifecycle. With more than 11,000 employees and clients operating across multiple industries, Endava has treated itself as "client zero," testing AI internally before advising other companies about how to introduce it across their operations.

Matt shares the story of a CEO who proudly told him that his company had completed its AI transformation after purchasing 10,000 licenses for an AI tool. Twelve months later, the business had seen little return on its investment and returned for help understanding what becoming AI-native actually required. The story captures one of the biggest problems with enterprise AI adoption today: buying technology is easy, but changing how people think about problems, redesign workflows, and create business value is much harder.

We discuss why Matt believes becoming AI-native is primarily a mindset. Rather than treating AI as another application added to the technology stack, employees should become curious about where AI can improve existing processes, remove unnecessary work, and create new ways of delivering value.

Matt also explains his idea that AI works best when it becomes invisible. Instead of requiring employees to constantly interact with chatbots and standalone AI applications, software agents can operate inside existing workflows, monitor information, prepare responses, identify problems, and bring people into the process when human judgment is required.

His own use of AI agents provides a practical example. While attending meetings that prevented him from monitoring email for several days, Matt used agents to review incoming messages, redirect requests, identify urgent communications, and prepare draft responses. Rather than handing complete control to automation, he determined which actions required approval and where AI could operate independently.

This leads to a wider discussion about human oversight and accountability. Matt argues that managing AI agents may increasingly resemble managing teams. Leaders do not inspect every decision made by every employee, but they establish responsibilities, controls, escalation points, and circumstances where intervention is required. Companies introducing agentic AI need similar approaches to supervision.

We also examine two mistakes Matt frequently sees companies make. The first is treating AI adoption as a software rollout, buying tools for employees and expecting productivity gains to appear automatically. The second is creating centralized AI centers of excellence and expecting a small group of specialists to determine how every department should use the technology.

Matt argues that employees closest to business processes are often best placed to identify opportunities for improvement. At Endava, the legal team runs monthly AI hackathons to redesign its own workflows, supported by technology specialists but led by people who understand the work itself.

For companies operating in payments, financial services, and other regulated industries, the conversation turns to reliability, auditability, traceability, and risk. Matt explains how Dava.Flow allows companies to translate regulatory requirements and operational controls into policies that AI systems must follow and demonstrate throughout the delivery process.

Rather than searching for a single killer AI application, Matt recommends examining end-to-end business workflows. Companies can map how information moves between employees, departments, and systems, identify unnecessary handoffs and manual processes, and determine where AI agents can improve speed, cost, and performance without replacing entire technology platforms.

Leadership is another major theme throughout the episode. Matt believes the companies that achieve meaningful results from AI will be led by executives who personally use the technology, understand its capabilities, and demonstrate the behaviors they expect from their workforce.

He shares how Endava brought senior leaders from legal, technology, people, and other business functions together to build software agents themselves. The experience changed how executives thought about technology investments, including one leader realizing that an existing vendor contract might no longer be necessary because the company could build the required capability internally.

For CIOs, CTOs, technology leaders, and business executives under pressure to demonstrate returns from AI investment, this conversation provides practical lessons on becoming AI-native, redesigning workflows, managing software agents, maintaining human accountability, operating AI in regulated industries, and moving beyond technology adoption toward measurable business value.

The companies that succeed with AI may not be those buying the most tools or making the biggest announcements. They will be the ones whose leaders understand the technology, whose employees rethink how work gets done, and whose AI systems quietly become part of everyday business operations.